diff options
Diffstat (limited to 'results/scraper/launchpad/1860610')
| -rw-r--r-- | results/scraper/launchpad/1860610 | 20 |
1 files changed, 20 insertions, 0 deletions
diff --git a/results/scraper/launchpad/1860610 b/results/scraper/launchpad/1860610 new file mode 100644 index 00000000..0fc4f748 --- /dev/null +++ b/results/scraper/launchpad/1860610 @@ -0,0 +1,20 @@ +cap_disas_plugin leaks memory + +Looking at origin/master head, the function cap_disas_plugin leaks memory. + +per capstone's examples using their ABI, cs_free(insn, count); needs to called just before cs_close. + +I discovered this running qemu under valgrind. + +It looks like this will fail on all the other capstone cases as well. Is this an API change across versions? + +I run git blame in the capstone repository, and cs_free has been around for at least 4 years in the capstone ABI. I can not tell if the need to call cs_free is a (new) requirement. Documentation capstone is a little informal... + +What command line where you using? I've been unable to replicate the valgrind warning with a riscv64-linux-user run of hello with the libhowvec.so plugin. Valgrind does complain about a bunch of other stuff though. + +Looking at the way disas is structured it seems cap_insn is allocated once (per thread) and re-used for each disassembly so we shouldn't be free'ing it after each usage. In fact the comments to cap_disas_start imply we want to do better than re-initialising the library for every set of instructions we disassemble. + +It is true that we don't clean-up any of the disassembly machinery on exit but the same can be said for a lot of QEMU's static state. So currently I don't see a leak rather than a one-time allocation. Unless I can reproduce the leak I'm going to mark this as incomplete for now. + +[Expired for QEMU because there has been no activity for 60 days.] + |